Mycroft and SNIPS: open source voice assistants

Any excuse to have Sherlock on the blog is good, I say. This time I want to record the fact that there is an open source voice assistant, called Mycroft, code available from GitHub, originating from Kansas City, MO, they say. Mycroft, like Alexa and Bixby, has skills contributed by the community of users. I wanted to buy Mycroft Mark 2, but this seems to be out of stock right now.

The other open source voice assistant is called snips. They have a different storyline in mind and Jarvis is the module that creates a voice assistant for free, in different languages. Then one can deploy the assistant to a Raspberry Pi, an Android or a Linux device. This will run completely on-device, keeping data safe and private! 
The Snips Platform is a software solution powering Private-by-Design voice assistants. Businesses or individuals, anyone can set up the Snips Platform on a single board computer (for example a Raspberry Pi 3, an i.MX8M board, an Android or an iOS device), and install a voice assistant on it.

Alexa has by now apparently 70K skills and I don't know how many `tasks' in Google Home Assistant market place. I wonder if anyone has written a comparison of all these voice frameworks.

Natural Language Inference: over and over again

This post is about collecting the many talks I have given about Natural Language Inference.
The first one is from the beginning of 2017 and has not been written up as a paper, yet. The slides of Universal Dependencies for Inference from Feb 2017 are in SlideShare.

Then the work with Katerina Kalouli and Livy Real, that ended up in several papers, is described in the slide deck Natural Language Inference: logic from humans, from December 2017.

Further,  I talked in Oslo on Natural Language Inference in SICK at the  MAy 2018 workshop on Meaningful Work: Advancing Computational Semantics. There I met Martha Palmer and more collaboration ensued, yay!